Seung Il Lee

I'm Seung Il Lee. My research focuses on multimodal language models and AI agents. I am interested in leveraging multimodal models to manage and control intelligent agents.

Email  /  Github  /  Blog  /  Tistory

Publications

We propose a token-based affordance grounding framework using Large Vision-Language Models (LVLMs) to localize action-supporting regions in an image.

📰 Featured in The Conversation: "From lab to life: Why translating AI advances to the real world is a major challenge"

Authored an introductory technical guide exploring the core principles of artificial intelligence through foundational research papers, tracing the key questions, intuitions, and experimental blueprints behind models like ChatGPT and generative AI.

Authored a technical book covering Web 3.0 implementation, blockchain network architecture, and hands-on Python development.

Professional Experience

• Global AI Product Launch (NYSE): Architected the AI-driven investment logic and container-based infrastructure (AWS) for the GXIG ETF / news, successfully listed on the NYSE.
• Cloud-Native Collaboration: Established a global collaborative development environment using Git and AWS EKS, co-working with New York-based teams to ensure system consistency.
• AI Asset Management: Designed and built an AI-based wealth management system, automating portfolio rebalancing and risk assessment using Python and Spark.
• Internal Process Automation: Developed an LLM-based automation tool for internal employee and client management on GCP and AWS, reducing manual operational tasks.

• Enterprise LLM Agent: Built and deployed an LLM-based financial data agent serving 3 million retail customers / news.
• Real-time AI Pipeline: Engineered a high-performance data pipeline using LangChain and Kafka to ingest and process unstructured data from Reuters, CNBC, and SEC EDGAR.
• Large-scale Data Analysis: Utilized PySpark to analyze behavioral data of 3 million customers, providing personalized investment insights through AI models.
• Scalable RAG Architecture: Developed a production-grade RAG pipeline to ensure high-accuracy information retrieval from real-time global financial news feeds.

• Real-time IoT Pipeline: Built a Kafka-based streaming data pipeline on AWS to process TB-scale real-time vehicle sensor data.
• Driving Risk Analytics: Developed algorithms to analyze driving patterns and measure safety risks, providing foundational data for insurance scoring models.
• Serverless Architecture: Optimized cloud costs and performance by implementing serverless components (Lambda, Athena) for geospatial data analysis.

• Operational Performance (OEE): Researched and designed the OEE analytical framework from first principles to optimize semiconductor manufacturing efficiency.
• Predictive Risk Modeling: Developed statistical "Down-risk" models to identify equipment failure patterns, reducing unplanned downtime in the production line.
• Data-driven Decision Support: Translated complex manufacturing logs into structured KPIs for executive-level strategic production planning.

• Data-driven Strategy: Established mid-to-long term business strategy for Kia Motors based on data analysis.

Honors & Awards

• First Prize in U+ Customer Idea Contest | LG Uplus | 2022
• First Prize in MyData Contest | Ministry of the Interior and Safety (MOIS) | 2022
• Grand Prize for Blockchain Challenge | Korea Internet & Security Agency (KISA) | 2022
• Excellence Award in News Big Data Hackathon | Korea Press Foundation | 2022
• Award for Production Optimization through Equipment Log Analysis | Samsung Electronics | 2018

Academic Services

• Reviewer: NeurIPS 2026

Feel free to steal this website's source code.